Self-adaptive Software Hepu And Application Research

Self-adaptive software can change software behavior autonomously and dynamically with the changing software operation environment.In application,resource allocation is an important part of software.However,at present,resource allocation of application cannot be adjusted flexibly and dynamically according to software status,when it is adjusted according to the established strategy of developers.Therefore,it is of great practical significance to study the self-adaptive software.This thesis proposes a universal solution for adaptive software-Hepu combined with resource allocation scenarios.First of all,the Hepu software system model is proposed,which solved the problem of high coupling of traditional self-adaptive software and combines the agent-environment architecture of reinforcement learning technology,Secondly,according to different application scenarios and domain models,the environment acquisition component model,environment analysis component model and environment deployment component model of Hepu are established.Thirdly,two agent decision algorithm are designed to solve the problem of over-relevance of application data,one is based on a generative confrontation network and another is based on parallel training.Fourthly,Hepu was designed and developed according to software requirements.Finally,the establishment of virtual environment for network resource allocation and cache resource allocation of the icon social software have verified the adaptive effect of Hepu application in the resource allocation scenario.Hepu is an important tool,which not only implements adaptive resource allocation in icon social software,but can also be extended to other application.